Spring, 2005 (Roos, Soc. 502)

Assignment 3: Interacting with Dummies (due Wednesday, March 2nd)

This assignment is an attempt to get you a bit closer to a finished model for your final paper, as well as to use dummy variables and interaction terms. Your final paper for the course will be a quantitative analysis of a substantive issue of your own choosing. The purpose is to show to good advantage the technical and analytical skills you are developing (always, of course, within a theoretical framework).

You should also begin to review the literature in the area you've chosen. By the end of the semester you should have a fairly sophisticated analytic review of works relevant to your topic, to frame your quantitative work. If you plan to turn your paper into a dissertation chapter or paper to present at professional meetings, the more you do now about researching the extant literature, the better off you will be. The major reason for doing as much of this as possible now is that previous literature can indicate potential variables of interest that ad hoc theorizing might not. Next week Jeris Cassel, a social science librarian, will go over library search techniques to get you going.

For Assignment 3, estimate a multiple regression model that includes the variables that you hypothesize affect your dependent variable. Include at least one set of dummy variables. [Don't forget to create a dummy variable for each category of the nominal variable, but only include n-1 in the regression model.] Follow the procedure we talked about in class (calculating a set of dummy variables and associated interaction terms, and estimating Models 1, 2, and 3) to determine whether separate equations for the categories of the nominal variable are justified. This will tell you whether the groups do in fact have significantly different slopes from the reference category. [Hint: choose a dummy variable (or a dichotomous variable) of theoretical importance to you to be the variable you add in for Model 2, and for which you create interaction terms. For example, if you think males and females differ in their average values on the dependent variable, and that their slopes differ significantly as well, then sex should be the variable you add in for Model 2 and that you interact with all other independent variables.]

As usual, write up the discussion of your results as if you were writing a journal article. At the minimum, you should have at least the following tables: (1) a table that gives the unstandardized regression results for Models 1, 2, and 3, and an accompanying table showing means and standard deviations (don't include correlations, they are overkill given the number of variables you will have); (2) if you find separate equations justified, you should go on to present a table with means, standard deviations, and correlations for the separate models (i.e., male vs. female) and another table showing your regression results. Your final paper may, of course, include additional tables that describe your coding decisions, provide descriptive overviews of your sample, relevant crosstabs, and so forth. Focus on describing the substantive meanings of your numbers. Turn in your final log and output ONLY if you think you might be having problems.